Structural change and New Zealand s productivity performance

Transcription

1 Structural change and New Zealand s productvty performance Workng Paper 214/4 June 214 Author: Lsa Meehan

2 New Zealand Productvty Commsson Workng Paper 214/4: Structural change and New Zealand s productvty performance Date: June 214 Author: Lsa Meehan JEL classfcaton: L16; O46; O57; O56 JEL keywords: Industry Mx; Macroeconomc Industral Structure; Productvty; New Zealand; Comparatve Country Studes. ISBN: (onlne only) Acknowledgements: Thanks to New Zealand Productvty Commsson staff who provded useful comments and suggestons, partcularly Paul Conway (Drector Economcs & Research) and the Servces Inqury team led by Geoff Lews. Also thanks to staff of the New Zealand Treasury, Mnstry of Busness, Innovaton & Employment and Statstcs New Zealand who nputted nto ths analyss, ncludng Roger Proctor, John Janssen and Adam Tpper. I am grateful to Statstcs New Zealand for assstance wth the New Zealand data. Dsclamer: The vews expressed n ths Workng Paper are strctly those of the author. They do not necessarly reflect the vews of the New Zealand Productvty Commsson or the New Zealand Government. The author s solely responsble for any errors or omssons. Informaton on the Productvty Commsson can be found on or by contactng

3 Structural change and New Zealand s productvty performance 3 Abstract Shft-share analyss decomposes aggregate labour productvty growth nto a contrbuton from wthnndustry productvty growth and a contrbuton from employment movements across ndustres wth dfferng labour productvty levels. Because the role that structural change plays n productvty growth dffers wth the level of a country s economc development, ths paper focuses on New Zealand n comparson wth other OECD countres. New Zealand s economy-wde labour productvty growth has been lower than most other OECD countres. As n all other OECD countres examned n ths paper, the majorty of New Zealand s labour productvty growth snce the early 199s has come from wthn-ndustry productvty growth. Lke most other OECD countres, New Zealand has experenced productvty-detractng structural change wth employment movng towards ndustres wth below-average levels of labour productvty. In aggregate, New Zealand s poor labour productvty growth compared wth other OECD countres reflects both below-average performance of ts wthn-ndustry productvty growth and a larger employment shftng towards low-productvty ndustres. Although the structural change component s a relatvely smaller part of overall labour productvty growth, New Zealand s structural change effect was further behnd the OECD average structural change effect than ts wthn-ndustry productvty growth. New Zealand s comparatvely large negatve structural change was due to small dfferences n employment share changes and relatve labour productvty levels n a few ndustres. Industres that were undergong sgnfcant reforms durng the 199s were the man culprts, wth larger employment movements away from hgh labour productvty ndustres such as electrcty, gas & water and transport, storage & communcatons n New Zealand than other OECD countres. These dfferences hghlght some of the lmtatons of shft-share analyss that may be partcularly acute durng tmes of reform. For example, the assumpton that the average and margnal productvty n an ndustry are equvalent may result n an overestmaton of the negatve contrbuton of structural change. As a small open economy, New Zealand could, n prncple, ncrease output n hgh-productvty ndustres through exportng. In practce, New Zealand s propensty to export s low compared wth other small, open economes and ths export-led employment shft has not occurred. The shft-share analyss presented n ths paper s a frst step n understandng the role of resource reallocaton n productvty growth n New Zealand. In nterpretng the results of ths paper t s mportant to note that shft-share analyss s a descrptve tool to look at hgh-level structural change. Taken n solaton, t does not provde nformaton on the drvers and dynamcs of resource reallocaton, nor does t necessarly ndcate the ms-allocaton of resources. Frm-level data offers the opportunty to explore these ssues further.

4 Contents Abstract... 3 Contents Introducton Methodology and data Internatonal comparsons Detaled New Zealand analyss Internatonal comparsons Structural transformaton and productvty growth Shft-share analyss: nternatonal comparsons A closer look at New Zealand Shft-share n detal Dscusson Interpretaton and the role of reform Increasng output n hgh-productvty ndustres through exportng Concluson References Appendx A Shft-share methodology n detal Appendx B Estmatng hours pad for New Zealand Appendx C Statstcs New Zealand ndexes versus calculated growth rates Tables Table 1 Summary of ndustry contrbuton to structural change level and growth effects for detaled New Zealand analyss... 9 Fgures Fgure 1 Labour productvty growth shft-share decomposton, OECD countres, Fgure 2 Shft-share decomposton as a share of labour productvty growth, OECD countres, Fgure 3 Economy-wde labour productvty growth versus the structural change effect, Fgure 4 Intal ndustry employment shares: New Zealand versus OECD average, Fgure 5 Industry labour productvty growth rates: New Zealand versus OECD average, Fgure 6 Industry contrbutons to the wthn-ndustry effect: New Zealand versus OECD average, Fgure 7 Industry contrbutons to the structural change effect: New Zealand versus OECD average, Fgure 8 Industry productvty and change n employment share: New Zealand versus OECD average, Fgure 9 Average ndustry employment shares for country groups, Fgure 1 Agrcultural productvty relatve to manufacturng productvty by country, Fgure 11 Change n economy-wde labour productvty f a country had average OECD employment structure, Fgure 12 Shft-share decomposton of labour productvty growth: Asa and Latn Amercan countres,

5 Structural change and New Zealand s productvty performance 5 Fgure 13 Shft-share decomposton of labour productvty growth over productvty cycles, Fgure 14 Industry contrbutons to labour productvty growth between 1978 and Fgure 15 Industry contrbutons to labour productvty growth over productvty cycles n the 199s and 2s Fgure 16 Hours pad and labour productvty growth n selected ndustres Fgure 17 Growth n employment (hours) n OECD countres, Fgure 18 Average labour productvty and captal share of ncome, Fgure 19 Industry productvty and change n employment share: Australa,

6 1 Introducton The Terms of Reference for the New Zealand Productvty Commsson s nqury nto Boostng Servces Sector Productvty asks for analyss on the extent to whch employment has shfted from hgh to low productvty sectors. 1 To address ths Terms of Reference tem, ths paper uses shft-share analyss to decompose labour productvty growth nto contrbutons from wthn-ndustry productvty growth and structural change. The concern over the movement of employment towards low-productvty ndustres stems from recent lterature hghlghtng the mportance of structural change n drvng productvty growth n developng countres. In partcular, McMllan & Rodrk (211) fnds that dfferences n structural change have been more mportant than wthn-ndustry productvty growth n explanng dfferences n the labour productvty performance of Asan and Latn Amercan countres. Asan countres have generally experenced productvty-enhancng structural change as employment has moved from low- to hghproductvty ndustres whle employment has moved towards low-productvty ndustres n Latn Amercan countres. A country s stage of economc development matters for the role of structural change n productvty growth. Developng countres generally have greater potental for productvty-enhancng structural change as employment moves from low-productvty agrculture to hgher-productvty ndustres, partcularly manufacturng. The role of structural change n productvty growth s less clear n developed countres where employment tends to move from agrculture and manufacturng ndustres to servce ndustres, ncludng low-productvty communty and personal servces (dscussed n Secton 3.1). Whle much of the nternatonal lterature concentrates on developng countres, ths paper focuses on New Zealand n comparson wth other OECD countres. New Zealand s economy-wde labour productvty growth has been lower than most other OECD countres, wth ts growth between 199 and 25 rankng 14 th out of the 2 OECD countres examned n ths paper. The majorty of labour productvty growth n New Zealand, as n all the other OECD countres ncluded n the analyss, came from wthn-ndustry productvty growth. Lke most other OECD countres, New Zealand experenced productvty-detractng structural change wth employment movng towards ndustres wth below-average levels of labour productvty. New Zealand s comparatvely poor labour productvty growth reflects a below-average performance of ts wthn-ndustry productvty growth and a larger employment shft towards low-productvty ndustres. Although the structural change component was a smaller part of overall labour productvty growth, New Zealand s structural change effect was further behnd the OECD average than ts wthn-ndustry growth (Secton 3.2). A closer look at the ndustry contrbutons to New Zealand s comparatvely large negatve structural change effect shows t was due to small dfferences n employment share changes and relatve labour productvty levels n a few ndustres. Industres that were undergong sgnfcant reforms durng the 199s were the man culprts, wth larger employment movements away from hgh labour productvty ndustres such as electrcty, gas & water and transport, storage & communcatons n New Zealand than other OECD countres (Secton 3.2). A more detaled look at New Zealand confrms the negatve contrbutons to the structural change effect of these reformng ndustres (Secton 4). These dfferences hghlght some of the lmtatons of shft-share analyss that may be partcularly acute durng tmes of reform (Secton 5.1). For example, the assumpton that average and margnal productvty of an ndustry are equvalent may result n an overestmaton of wthn-ndustry productvty growth component and an overestmaton of the negatve contrbuton of structural change (Tmmer & Szrma, 2). 1 The full Terms of Reference and Inqury reports are avalable from

7 Structural change and New Zealand s productvty performance 7 As a small open economy, New Zealand could, n prncple, ncrease output n hgh-productvty ndustres through exportng. In practce, ths export-led employment shft has not occurred, and growth n real non-tradable output has outstrpped growth n real tradable output snce the md-2s (New Zealand Treasury, 212). New Zealand s propensty to export s low compared wth other small, open economes, possbly reflectng the hgh fxed costs of exportng faced by New Zealand frms (Secton 5.2). The shft-share analyss presented n ths paper s a frst step n understandng the role of resource allocaton n productvty growth n New Zealand. In nterpretng the results of ths paper t s mportant to note that shft-share analyss s a descrptve tool to look at hgh-level structural change. Taken n solaton, t does not provde nformaton on the drvers and dynamcs of resource reallocaton, nor does t ndcate the ms-allocaton of resources. Frm-level data offers the opportunty to explore resource reallocaton and productvty growth n more detal, and has the potental to provde nsghts nto the drvers and dynamcs of resource reallocaton, ncludng the role of polcy. 2 Methodology and data Shft-share analyss decomposes aggregate labour productvty growth nto wthn-ndustry productvty growth and the contrbuton of structural change. Although the method has several varants, the basc dea nvolves algebrac rearrangement of the labour productvty growth equaton. 2.1 Internatonal comparsons The nternatonal comparson work presented n Secton 3 below apples (almost) the same equaton as Rodrk & McMllan: P = S P + ( P P) S t, k t, t where P s the change n aggregate labour productvty, St, k s ndustry s ntal share of aggregate employment, P s the change n ndustry s labour productvty level between tme t-k and t, S s the change n ndustry s share of employment over the same perod, P t, s ndustry s end-perod level of labour productvty and Pt s aggregate end-perod labour productvty. The dervaton of Equaton 2.1 s detaled n Appendx A. In words, ths equaton decomposes labour productvty growth nto a wthn-ndustry effect and a structural change effect. The wthn-ndustry effect s the weghted sum of ndustry labour productvty growth, where the weghts are the ntal ndustry employment shares. The structural change effect s the ndustry s labour productvty level at the end of the perod (relatve to aggregate productvty) multpled by the ndustry s employment share change. The structural change effect s postve f employment has moved towards ndustres wth above-average levels of labour productvty. In applyng ths methodology, the nternatonal comparsons n Secton 3 use the November 29 EUKLEMS database, whch allows the shft-share framework to be appled on a comparable bass across 19 OECD countres. 3,4 The same tme perod (199 to 25) as McMllan & Rodrk s used, but 2 The only dfference between ths equaton and McMllan & Rodrk s s that the structural change effect used here measures ndustry labour productvty relatve to aggregate labour productvty. Whle ths dfference has no mpact on the aggregate reallocaton effect, t makes the ndustry contrbutons to the structural change effect easer to nterpret. 3 The EUKLEMS database s avalable from 4 Ths paper uses the Statstcs New Zealand conventon of referrng to ANZSIC6 1-dgt categores (or roughly equvalent ndustry categores) as ndustres and a group of ndustres as a sector. For example, the prmary sector s made up of the agrculture, forestry & fshng ndustry and the mnng ndustry.

8 a greater degree of ndustry dsaggregaton s used (14 ndustres nstead of nne). 5 Whle McMllan & Rodrk measure employment as the number of workers, ths paper uses hours worked. 6 The EUKLEMS database does not nclude New Zealand. It s added by makng Statstcs New Zealand data as close as possble to the data for other countres. As such, some cauton should be appled n comparng New Zealand to other countres due to remanng data comparablty ssues. For example, the New Zealand data uses hours pad as the labour nput measure because hours worked are not avalable at the ndustry level for New Zealand, 7 and the ndustry classfcatons used n the New Zealand data do not algn exactly wth the ndustry classfcatons n the EUKLEMS database. Hours pad n some ndustres are not avalable for New Zealand and were estmated (see Appendx B for detals). Although ths paper focuses on comparng New Zealand wth other OECD countres, data from the GGDC 1-sector database was used to also brefly look at eght Latn Amercan countres and eght Asan countres. Due to data avalablty, ths analyss measures employment usng the number of workers and uses a hgher level of ndustry aggregaton than the analyss for OECD countres. Whle cross-country analyss provdes useful broad comparsons, t s prudent not to read too much nto small dfferences among countres. A lmtaton of shft-share analyss s that t uses data from two ponts n tme, and s therefore senstve to the choce of the start- and end- perod. The tmng of busness cycles across countres dffers, whch may mpact on the cross-country comparablty partcularly as busness cycles wll have a dfferent mpact on dfferent ndustres. For example, f employment n low-productvty servce ndustres, such as personal servces and retal trade, s more senstve to the busness cycle, shft-share analyss may overestmate the negatve structural change effect of a country that was n a contractonary phase n 199 and an expansonary phase n Detaled New Zealand analyss Secton 4 examnes the New Zealand experence n more detal than the nternatonal comparsons secton. It splts the structural change effect nto two terms: a level effect and a growth effect followng a varant of the shft-share equaton used n Sharpe (21): S P ( P P ) S ( P P) S P = + + P P P P, t k, t k t k t k t k t k t k 2.2 That s, labour productvty growth s decomposed nto a wthn-ndustry effect (the frst term on the rght-hand sde of Equaton 2.2), a structural change level effect (the second term) and a structural change growth effect (the thrd term). The level effect s postve f labour has moved towards ndustres wth above-average levels of productvty and the growth effect s postve f labour has moved towards ndustres wth above-average labour productvty growth. Appendx A detals the dervaton of Equaton 2.2. The New Zealand-specfc analyss gves more detals of the contrbuton of ndvdual ndustres to the overall results, partcularly the contrbuton of ndustres to the structural change effect. It s therefore useful to note the dfferent ways n whch an ndustry can make a postve or negatve contrbuton to the wthn-ndustry effect and the structural change level and growth effects. The drecton of 5 McMllan & Rodrk use nne ndustres because they examne more countres usng the less detaled GGDC 1-sector database. The focus here s on the OECD countres ncluded n the EUKLEMS database, and therefore, the greater level of detal s used. EUKLEMS also has data for earler years, however employment data by ndustry for New Zealand could not be estmated further back than Hours worked are unavalable for many of the non-oecd countres examned n McMllan & Rodrk. 7 Although t would be preferable to use the number of hours worked n an ndustry to take nto account factors such as pad leave and unpad overtme, the ndustry hours pad seres are consdered more robust (see Statstcs New Zealand, 213b for detals). 8 For a full dscusson of the lmtatons of shft-share analyss, see Tmmer & Szrma (2).

9 Structural change and New Zealand s productvty performance 9 contrbuton s smple for the wthn-ndustry effect, as an ndustry makes a postve contrbuton f labour productvty growth for that ndustry s postve. An ndustry can make a postve contrbuton to the structural change level effect n one of two ways. Frst, t can have a hgher-than-average level of labour productvty and account for a growng share of employment (e, P, t k Pt k > and S > ). Or t can have a lower-than-average level of productvty and account for a decreasng share of employment (e, P P < and S < ). Industres that have ether low labour productvty and an ncreasng employment share, or hgh labour productvty and a decreasng employment share, wll make a negatve contrbuton to the level effect. Ths s summarsed n Table 1. An ndustry can make a postve contrbuton to the structural change growth effect f ts productvty growth s above average and ts share of employment s growng (e, P P > and S > ) or ts productvty growth s below average and ts share of employment s decreasng (e, P P < and S < ). 9 It wll make a negatve contrbuton f t has above-average productvty growth but accounts for a decreasng share of employment or f ts productvty s below-average and t accounts for an ncreasng share of employment (Table 1). The magntude of an ndustry s contrbuton to the three rght-hand-sde components of Equaton 2.2 wll depend on ts sze or change n sze and ts relatve level of labour productvty or relatve labour productvty growth. Table 1 Summary of ndustry contrbuton to structural change level and growth effects for detaled New Zealand analyss Level effect Decreasng share of employment S < Increasng share of employment S > Low productvty levels relatve to aggregate P, P < t k t k Postve contrbuton Negatve contrbuton Hgh productvty levels relatve to aggregate P, P > t k t k Negatve contrbuton Postve contrbuton Decreasng share of employment S < Increasng share of employment S > Growth effect Low productvty growth relatve to aggregate P P < Postve contrbuton Negatve contrbuton Hgh productvty growth relatve to aggregate P P > Negatve contrbuton Postve contrbuton Statstcs New Zealand data are used to apply ths methodology. The man dfferences between the nternatonal comparsons data and the New Zealand data are: ndustry coverage, the degree of ndustry dsaggregaton and the tme perod examned. The New Zealand results focus on the 11 market-sector ndustres from 1978 to 211. These 11 ndustres cover about 6% of hours pad and 66% of current prce GDP n 21 and exclude hard-tomeasure servce and government-domnated ndustres. Whle t would be preferable to examne the 9 Note that ths s not a comparson of ndustry productvty growth rates wth the aggregate growth rate. Rather, an ndustry s absolute change n labour productvty s compared wth the absolute change n aggregate labour productvty. For example, an ndustry wth a low level of labour productvty may experence a larger-than-average ncrease n growth rate terms due to the low startng pont, but the absolute change may stll be smaller than the aggregate absolute change.

10 whole economy, the choce of 11 ndustres s a trade-off favourng the longer tme coverage and robustness of the 11-ndustry-market-sector seres over fuller coverage of economc actvty. The ndustres used n the New Zealand secton are more narrowly defned than those used n the nternatonal secton. The analyss, however, stll nvolves a hgh degree of ndustry aggregaton. For example, manufacturng s not dsaggregated nto sub-ndustres. Ths hgh degree of aggregaton may lead to an underestmaton of the mpact of structural change as t does not capture, for example, the movement of employment from low-productvty manufacturng sub-ndustres to hgh-productvty ones. Whle Statstcs New Zealand does not release more dsaggregated labour volume seres, a possble extenson of ths work could nvolve usng the dataset of Mason (213) whch has manufacturng hours data dsaggregated nto nne sub-ndustres. Another possble and very worthwhle extenson s the use of Statstcs New Zealand s Longtudnal Busness Database to look at the contrbuton of employment reallocaton at the frm level. 1 The results n Secton 4 are presented over the whole 1978 to 211 perod and over aggregate productvty growth cycles. Breakng the data nto cycles better accounts for factors that tend to vary wthn a cycle, such as capacty utlsaton (Statstcs New Zealand, 27). Aggregate productvty cycles are, however, unlkely to completely elmnate cyclcal components as ndustry cycles may not match economy-wde cycles. 11 There are sx complete cycles over the 1978 to 211 perod: 1978 to 1982, 1982 to 1985, 1985 to 199, 199 to 1997, 1997 to 2 and 2 to 28. The latest perod, 28 to 211, s an ncomplete cycle. 3 Internatonal comparsons 3.1 Structural transformaton and productvty growth Ths paper focuses on New Zealand n comparson wth other OECD countres. As mportant background, however, ths sub-secton outlnes the general path of structural transformaton and hghlghts the dfferent role structural change plays n the productvty growth of developng versus developed countres. The general path of economc development nvolves resources movng from agrculture, to manufacturng, and then to servce ndustres. In the early stages of economc development, agrculture employs the majorty of workers. As the economy ndustralses, employment n agrculture decreases and employment shfts towards manufacturng. In later stages of development, employment shfts from agrculture and manufacturng ndustres to the servces sector (Kruger, 28). Most recent lterature on structural change and ts role n productvty growth focuses on developng countres, reflectng the greater scope for productvty-enhancng structural change n these countres. 12 Generally, a developng country not only has a larger share of employment n agrculture, but ts labour productvty n agrculture s consderably lower than n developed countres (both n absolute terms and relatve to the productvty of a country s other ndustres). As an economy develops, excess labour moves from agrculture to hgher-productvty ndustres, such as manufacturng, resultng n productvty-enhancng structural change. McMllan & Rodrk (211) hghlghts the mportance of structural change to productvty growth n developng countres. Snce 199, developng Asan countres have generally experenced productvty-enhancng structural change, wth employment movng from low-productvty agrculture 1 The Productvty Hub plans on dong ths frm-level work as part of ts forward-lookng research agenda. See Nolan (214) detals. 11 Industry productvty cycles have not been calculated for New Zealand. Australan research suggests some ndustry cycles, such as agrculture, may dffer from the aggregate productvty cycle (Barnes, 211). 12 For example, McMllan & Rodrk (211), Tmmer & Szrma (2) and de Vres, Erumban, Tmmer, Voskoboynkov, & Wu (212).

11 Structural change and New Zealand s productvty performance 11 to hgher-productvty ndustres such as manufacturng. Employment n Latn Amercan countres, n contrast, has moved away from manufacturng and towards lower-productvty servce ndustres. McMllan & Rodrk suggest that the strong labour productvty growth n Asan countres versus the relatvely weaker growth n Latn Amercan countres s due not to dfferences n wthn-ndustry productvty growth, but to dfferences n the productvty mpact of structural change (see Box 1 for a dscusson of structural change n Asan and Latn Amercan countres). The potental contrbuton of structural change to productvty growth n developed countres s less clear. Developed countres do not tend to have large agrcultural ndustres wth excess labour. In addton, the varaton n labour productvty across ndustres tends to be lower than n developng countres. These factors leave less scope for productvty-enhancng structural change (at least n terms of resource allocaton at the ndustry level). 13 Whle developng countres generally mprove productvty through ndustralsaton, other forces may be mportant n developed countres. The share of employment n servces n post-ndustralsed economes has been rsng, ncludng employment n low-productvty labour-ntensve servces such as accommodaton, restaurants and recreatonal servces. Ths ncrease n employment n low-productvty servces may reflect a combnaton of Baumol s dsease and an ncrease n demand for these servces. Baumol s dsease means that labour-ntensve servce ndustres wll account for an ncreasng share of employment as the potental for productvty growth n these ndustres s low compared wth other ndustres (Baumol & Bowen, 1966). At the same tme, the demand for the output of these labourntensve servce ndustres tends to rse as ncomes rse (Denns & İşcan, 29). Indeed, the lack of consderaton of the demand-sde factors behnd resource movements s a lmtaton of shft-share analyss that flows from ts use of productvty, whch s a supply-sde concept (see Australan Productvty Commsson, 213 for a dscusson). The dfferences n ntal economc development and ndustry structure and the amount of varaton n ndustry labour productvty levels between developng and developed countres s borne out n the data (see Box 1). Snce the stage of a country s economc development matters for the role of structural change n productvty growth, the next sub-secton focuses on comparng New Zealand wth other OECD countres. 3.2 Shft-share analyss: nternatonal comparsons New Zealand s labour productvty growth was below the OECD average between 199 and 25 (1.8% a year versus 2.4%), and ranked 14 th out of 2 countres (Fgure 1). Ths lower productvty growth was due to a combnaton of below-average wthn-ndustry productvty growth and stronger movement of employment towards lower-productvty ndustres. The contrbuton of wthn-ndustry productvty growth was 2.2 percentage ponts n New Zealand compared wth the OECD average of 2.5 percentage ponts. In the majorty of countres (13 out of 2) employment moved towards ndustres wth below-average productvty, resultng n a negatve structural change effect. The average OECD structural change effect was -.1 percentage ponts. New Zealand s structural change effect was 17 th out of 2 at -.4 percentage ponts. As a percentage of aggregate labour productvty growth, New Zealand experenced the largest negatve structural change effect of all the examned OECD countres (Fgure 2). 13 The dstncton between resource reallocaton among ndustres and the reallocaton among frms wthn the same market s worth hghlghtng. There s an expectaton that n well-functonng markets, resources wll flow from low- to hgh-productvty frms wthn the same market, even f there s a less clear expectaton n developed countres that resources wll flow from low- to hgh-productvty ndustres.

12 Fgure 1 Labour productvty growth shft-share decomposton, OECD countres, % per year South Korea Ireland Sweden Fnland Unted Kngdom Portugal Japan OECD average Greece Germany Australa Unted States Austra France New Zealand Belgum Denmark Netherlands Luxembourg Span Italy Wthn-ndustry effect Structural change effect Labour productvty growth Source: EUKLEMS; Statstcs New Zealand; author s calculatons Fgure 2 Shft-share decomposton as a share of labour productvty growth, OECD countres, Greece Luxembourg Germany Italy France Australa Japan Sweden Portugal Austra Span Fnland Share of labour productvty growth (%) OECD average Belgum Denmark Unted States Ireland South Korea Netherlands Unted Kngdom New Zealand Wthn-ndustry effect Structural change effect Source: EUKLEMS; Statstcs New Zealand; author s calculatons The relatonshp between total labour productvty growth and the structural change effect s not clear. Some countres wth very hgh labour productvty growth had large negatve structural change effects (such as South Korea and Ireland) whle some countres wth low productvty growth had postve reallocaton effects (such as Italy and Luxembourg). If anythng, the relatonshp s slghtly downwards slopng (even f South Korea s excluded as an outler), wth countres experencng hgher labour productvty growth experencng greater negatve structural change effects (Fgure 3).

13 Structural change and New Zealand s productvty performance 13 Fgure 3 Economy-wde labour productvty growth versus the structural change effect, South Korea 6 Economy-wde labour productvty growth (% per year) Ireland Unted Kngdom Sweden Fnland 3 Portugal Japan Unted States Austra 2 New Zealand Denmark Belgum Netherlands 1 Span 5 4 Australa France Italy Germany Luxembourg Greece Contrbuton of structural change effect (percentage ponts) Source: EUKLEMS; Statstcs New Zealand; author s calculatons One ssue wth the precedng analyss s that t uses a hgh level of ndustry aggregaton. For example, Ireland s negatve structural change effect may be because manufacturng s not splt nto subndustres. If employment has shfted towards hgher-productvty manufacturng sub-ndustres n Ireland, the current shft-share analyss would attrbute ths movement of employment to wthnndustry productvty growth n manufacturng rather than a postve contrbuton to the structural change effect (and ths s lkely to be one of the reasons behnd Ireland s very hgh wthn-ndustry effect for manufacturng). Manufacturng s not the only pont of nterest n the Irsh reallocaton story. It had the largest employment share shft towards the low-productvty constructon ndustry n the OECD, consstent wth ts property boom. Ireland s constructon employment share ncreased by 7.1 percentage ponts between 199 and 25. Arguably, ths strong shft to constructon due to the property boom was a key reason behnd Ireland s unsustanable growth and subsequent poor economc performance n the second half of the 2s (OECD, 211). New Zealand s employment shft towards constructon was the ffth largest n the OECD (behnd Ireland, Span, Australa and Greece). Whle at 1.6 percentage ponts, New Zealand s employment share ncrease n constructon was much lower than Ireland s, unsustanable growth drven by a property and domestc-consumpton boom have also been rased as concerns for New Zealand (for example, see New Zealand Treasury, 21). It s also possble that New Zealand s comng constructon boom wll be larger than the boom of the early 2s as resdental housng growth n Auckland and the Canterbury rebuld drve unprecedented demand (Pacfecon (NZ) Ltd. & BRANZ, 213). It s dffcult to say whether New Zealand s structural change effect would be larger or smaller compared wth other OECD countres f a greater level of ndustry dsaggregaton was used. The most dsaggregated analyss nvolves frm-level nformaton, and assessng the contrbuton of frm ext, entry, wthn-frm productvty growth and reallocaton of resources across frms n the same ndustry s an mportant area of further research usng Statstcs New Zealand s Longtudnal Busness Database (Nolan, 214).

14 Industry contrbutons to the wthn-ndustry effect Whch ndustres contrbuted to New Zealand s 2.2% a year wthn-ndustry effect compared wth the OECD average of 2.5%? A New Zealand ndustry can make a lower contrbuton to the wthn-ndustry effect compared wth the OECD average f t had lower labour productvty growth and/or accounted for a smaller share of employment n 199. New Zealand had reasonably smlar ntal ndustry employment shares to the OECD average, albet wth relatvely large employment shares n agrculture (reflectng New Zealand s comparatve advantage n ths area); transport, storage & communcatons; fnance and real estate rental & busness servces, and relatvely smaller shares n manufacturng; constructon; wholesale & retal trade; publc admnstraton; healthcare and other servces (Fgure 4). The dfferences n contrbutons to the wthnndustry effects for New Zealand n comparson wth other OECD countres were, therefore, manly due to dfferences n ndustry productvty growth rates. New Zealand had much lower labour productvty growth rates n mnng; manufacturng; electrcty, gas & water and educaton than the OECD average, and hgher growth rates n agrculture; transport, storage & communcatons; healthcare and other servces (Fgure 5). 14 Fgure 4 Intal ndustry employment shares: New Zealand versus OECD average, 199 New Zealand OECD average % of total hours Agrculture Mnng Manufacturng Elec/gas/water Constructon Whsale/retal Htels/Rstaurants Trnsprt/storage/comms Fnance RealEst/Rent/Bus Publc admn Educaton Health Other servces Source: EUKLEMS; Statstcs New Zealand; author s calculatons The bggest contrbutor to New Zealand s lower wthn-ndustry effect was manufacturng (Fgure 6). The manufacturng ndustry n New Zealand had much lower labour productvty growth than the manufacturng ndustry n other OECD countres (2.% a year n New Zealand versus an OECD average of 5.3%) and also accounted for a smaller share of employment n 199 (18% n New Zealand versus 21% on average for the OECD). Wholesale & retal trade also made a smaller contrbuton to the wthn-ndustry effect n New Zealand than the OECD average due to lower ndustry labour productvty growth and a smaller share of employment. Agrculture; transport, storage & communcatons and fnance made larger postve contrbutons to the wthn-ndustry effect than the OECD average due to hgher labour productvty growth and larger shares of ntal employment n these ndustres n New Zealand. 14 The dfferences between labour productvty growth n New Zealand and other OECD countres n educaton and health care hghlght potental measurement ssues. Output s hard to measure for these ndustres, and s often assumed to ncrease proportonately wth nputs. New Zealand s one of the few countres that uses health and educaton outcomes (such as the number of qualfcatons awarded) to adjust output measures for these ndustres (see Statstcs New Zealand, 213a).

15 Structural change and New Zealand s productvty performance 15 Fgure 5 Industry labour productvty growth rates: New Zealand versus OECD average, % per year Agrculture Mnng Manufacturng Elec/gas/water Constructon Whsale/retal Htels/Rstaurants Trnsprt/storage/comms Fnance RealEst/Rent/Bus PublcAdmn Educaton Health Other servces OECD average New Zealand Source: EUKLEMS; Statstcs New Zealand; author s calculatons Fgure 6 Industry contrbutons to the wthn-ndustry effect: New Zealand versus OECD average, Agrculture Mnng Manufacturng Elec/Gas/Water Constructon Whsale/Retal Htels/Rstaurants Trnsprt/Storage/Comms Percentage pont contrbuton Fnance RealEst/Rent/Bus Publc admn Educaton Health Other servces OECD average New Zealand Source: EUKLEMS; Statstcs New Zealand; author s calculatons Notes: 1. Bars sum horzontally to the total wthn-ndustry effect n Fgure 1. Industry contrbutons to the structural change effect The man ndustres contrbutng to the larger negatve structural change effect n New Zealand than the OECD average were: agrculture; electrcty, gas & water; transport, storage & communcatons and real estate, rental & busness servces. Agrculture and real estate, rental & busness servces made a

16 smaller postve contrbuton to New Zealand s structural change effect than the average OECD contrbuton. Electrcty, gas & water supply and transport, storage & communcatons made larger negatve contrbutons than the OECD average (Fgure 7). Fgure 7 Industry contrbutons to the structural change effect: New Zealand versus OECD average, Percentage pont contrbuton Agrculture Mnng Manufacturng Elec/Gas/Water Constructon Whsale/Retal Htels/Rstaurants Trnsprt/Storage/Comms Fnance RealEst/Rent/Bus Publc admn Educaton Health Other servces OECD average New Zealand Source: EUKLEMS; Statstcs New Zealand; author s calculatons Notes: 1. Bars sum horzontally to the total structural change effect n Fgure 1. New Zealand s comparatvely large negatve structural change effect was due to relatvely small dfferences n changes n employment shares and relatve ndustry labour productvty levels compared wth the OECD average. Overall the employment share changes and relatve labour productvty levels n New Zealand are qute smlar to the OECD average (Fgure 8). Lke the average experence among OECD countres, manufacturng; agrculture; transport, storage & communcatons; utltes and mnng lost employment share n New Zealand, whle fnance; real estate & busness servces; constructon; wholesale & retal trade; hotels & restaurants and communty, personal & government servces ganed employment share. In addton, the same ndustres that have above-average labour productvty n New Zealand also have above-average labour productvty n the OECD on average, namely utltes; mnng; manufacturng; transport, storage & communcatons and fnance & busness servces. The smaller postve structural change effect from agrculture was due to a smaller decrease n employment share n New Zealand and hgher labour productvty relatve to the aggregate (Fgure 8). The employment share decreased by 3.6 percentage ponts n New Zealand between 199 and 25 versus 3.7 percentage ponts on average for the OECD countres examned, and New Zealand s agrculture productvty was 2% lower than New Zealand s aggregate labour productvty, versus an average of 4% lower than aggregate productvty for the OECD. The smaller postve structural change effect from real estate, rental & busness servces was due to a smaller ncrease n the ndustry s employment share n New Zealand compared wth the OECD average and a lower labour productvty level relatve to aggregate labour productvty. Real estate, rental & busness servces ganed 3.6 percentage ponts of the employment share n New Zealand, but an average of 4.8 percentage ponts for OECD countres. Labour productvty n the real estate, rental & busness servces ndustry was 31% hgher than New Zealand s aggregate labour productvty, but an average of 5% hgher for OECD countres.

17 Structural change and New Zealand s productvty performance 17 Fgure 8 Industry productvty and change n employment share: New Zealand versus OECD average, OECD average 2 Log of ndustry productvty/total productvty Manufacturng Elec/Gas/Water Mnng Fnance Trnsprt/Storage/ Comms Publc admn Whsale/Retal Educ Health Constructon Agrculture -.5 Other servces Htels/Rstaurants -1 Change n employment share, percentage ponts RealEst/Rent/Bus New Zealand 2 Log of ndustry productvty/total productvty 1.5 Elec/Gas/Water Fnance 1 Mnng Trnsprt/Storage/ Comms.5 RealEst/Rent/Bus Manufacturng Other servces Agrculture Health Publc admn Constructon -.5 Educ Whsale/Retal Htels/Rstaurants -1 Change n employment share, percentage ponts Source: EUKLEMS; Statstcs New Zealand; author s calculatons The larger negatve structural change effect for transport, storage & communcatons was due to a decrease n employment share for ths ndustry n New Zealand (-1.1 percentage ponts) versus a small shft n employment towards ths ndustry on average for the OECD (.4 percentage ponts) combned wth hgher relatve labour productvty levels for New Zealand. Transport, storage & communcaton s labour productvty levels were 14% hgher than New Zealand s aggregate labour productvty levels, versus an average of 39% hgher for the OECD countres. Electrcty, gas & water made a larger negatve contrbuton to the structural change effect n New Zealand than the OECD average due to a

18 larger decrease n employment share of ths ndustry n New Zealand whch was not fully offset by slghtly lower relatve labour productvty n New Zealand than the OECD average. New Zealand s employment share n ths ndustry decreased by.5 percentage ponts versus an average of.2 percentage ponts for the OECD, and ndustry labour productvty was 3.15 tmes hgher than the New Zealand aggregate, versus an average of 3.5 tmes hgher than the OECD. It s also worth reteratng that the comparson between New Zealand and other OECD countres s ndcatve only and small dfferences between New Zealand and other OECD countres should be nterpreted wth cauton. There are lkely to be comparablty ssues snce New Zealand s not ncluded n the EUKLEMS database and Statstcs New Zealand data was used to add New Zealand to the analyss. The relatvely large negatve structural change effect does, however, warrant further nvestgaton. Accordngly, the next secton takes a more detaled look at New Zealand. Box 1 Structural change and productvty growth n developng countres 15 Developng countres have greater potental for productvty-enhancng structural change. Ths greater potental stems from lower ntal levels of economc development, a hgh share of employment n lowproductvty agrculture, and a greater gap between productvty n agrculture and other ndustres, partcularly manufacturng. The most common measure of economc development, GDP per capta, was much lower n Latn Amercan and developng Asan countres than n New Zealand and other OECD countres. For example, New Zealand s GDP per capta n 199 was about three tmes hgher than that of Latn Amercan countres such as Costa Rca, Brazl and Chle, and more than eght tmes hgher than that of lower-ncome Asan countres such as Indonesa and the Phlppnes. A hgher startng share of employment n agrculture generally ndcates greater opportunty for employment to shft towards hgher-productvty ndustres. Latn Amercan and Asan countres had a much hgher share of employment n agrculture than OECD countres n 199. Recently-ndustralsed Asan countres, such as Tawan, had a larger share of employment than OECD countres n manufacturng, whle lower-ncome Asan countres such as Indonesa had a lower share of manufacturng employment. As expected, Asan countres had a lower average share of employment n communty, personal & government servces than OECD countres. Latn Amercan countres had, on average, a smlar proporton of employment n communty, personal & government servces as the OECD average, whch s hgher than expected gven the level of economc development (Fgure 9). 15 Due to data avalablty, the fgures n Box 1 measure employment as the number of workers (rather than hours worked) and ndustres are more aggregated here than n the man body of the paper. South Korea s ncluded n the Asa group of countres n Box 1, whereas t s ncluded n the OECD group n the man body of the paper. Mexco and Chle are ncluded n the Latn Amerca group of countres n Box 1 but s not ncluded n the OECD group n the man body of the paper due to lack of data on hours worked by ndustry. The fgures n Box 1, therefore, may not match exactly the fgures n the man body of the paper.

19 Structural change and New Zealand s productvty performance 19 Fgure 9 Average ndustry employment shares for country groups, % of total employment Asa Latn Amerca OECD Agrculture Constructon Wholesale, retal Fnance, busness servces Mnng, utltes Manufacturng Transport, communcatons Commuty, personsal, govt servces Source: GGDC 1-sector database; Statstcs New Zealand; author s calculatons Another element of the greater potental for productvty-enhancng structural change n developng countres s the larger gap between labour productvty n agrculture and other ndustres, partcularly manufacturng. The ntal gap between agrculture and manufacturng productvty was smaller n OECD countres than n Latn Amercan and Asan countres (Fgure 1). Fgure 1 Agrcultural productvty relatve to manufacturng productvty by country, 199 Agrculture productvty as % of manufacturng productvty OECD mean Argentna Colomba Costa Rca Latn Amerca mean Chle Mexco Venezuela Peru Brazl Malaysa South Korea Sngapore Asan mean (ex. HKG) Tawan Indonesa Phlppnes Latn Amerca Asa Source: Notes: GGDC 1-sector database; Statstcs New Zealand; author s calculatons 1. Hong Kong s excluded from ths graph and from the calculaton of the mean Asan fgures due to ts very hgh rato of agrculture productvty compared wth manufacturng productvty n 199 (about 188%). Another way to assess a country s potental to ncrease productvty through structural change s to examne how ts labour productvty would change f t retaned ts own ndustry productvty levels but

20 had OECD average employment shares. 16 For most OECD countres, adoptng the OECD average employment shares would have made relatvely lttle dfference because ndustry structure s reasonably smlar across OECD countres. Developng countres had more scope for productvtyenhancng structural change than OECD countres. Lower-ncome Asan countres n partcular had a great deal of scope for productvty-enhancng structural change. If Indonesa had the average OECD employment structure n 199 but retaned ts own ndustry labour productvty levels, ts aggregate labour productvty would have been more than twce as hgh (Fgure 11). Fgure 11 Change n economy-wde labour productvty f a country had average OECD employment structure, 199 % ncrease n economy-wde labour productvty Venezuela Argentna Chle Latn Amercan mean Colomba Mexco Brazl Costa Rca Peru Sngapore Hong Kong Malaysa Tawan South Korea Asa mean Phlppnes Thaland Indonesa Latn Amerca Asa Source: GGDC 1-sector database; author s calculatons Shft-share analyss shows the greater potental of Asan countres, partcularly lower-ncome Asan countres to mprove productvty through structural change compared wth developed countres. Labour productvty growth n all Asan countres except the Phlppnes was hgher than the OECD average growth rate between 199 and 25. Whle the structural change effect was on average about zero for OECD countres, the structural change effect n Asan countres was generally postve, wth employment movng towards hgher productvty ndustres (Fgure 12). The labour productvty growth of Asan countres was hgher than Latn Amercan countres, largely reflectng smlar wthn-ndustry productvty growth coupled wth generally postve structural change effects n Asan countres but negatve structural change effects n Latn Amerca (consstent wth McMllan & Rodrk s fndngs). Broadly speakng, whle employment has moved away from agrculture n both Asan and Latn Amercan countres, t has moved towards hgher-productvty manufacturng n Asan countres, but towards relatvely low-productvty servces n Latn Amercan countres. In ths respect, the employment movements n Latn Amercan countres look more lke those of developed countres, and the ntermedate stage of structural transformaton where employment from agrculture to manufacturng seems to have been bypassed. 16 Ths thought experment should not be taken too lterally as a country s unlkely to be able to modfy ts employment share wthout changng ts ndustry labour productvty levels.

21 Structural change and New Zealand s productvty performance 21 Fgure 12 Shft-share decomposton of labour productvty growth: Asa and Latn Amercan countres, OECD average Peru Chle Argentna Latn Amerca mean Costa Rca Mexco Brazl Colomba % per year Venezuela South Korea Malaysa Tawan Sngapore Asa mean Hong Kong Thaland Indonesa Phlppnes Latn Amerca Asa Wthn-ndustry effect Reallocaton effect Labour productvty growth Source: GGDC 1-sector database; author s calculatons 4 A closer look at New Zealand Ths secton takes a more detaled look at the role of structural change n New Zealand s labour productvty growth performance. As outlned n Secton 2, t uses a more detaled shft-share decomposton, a longer tme perod, more dsaggregated ndustres and measures labour nput as hours pad rather than the number of workers. It focuses on 11 market-sector ndustres rather than the whole economy, favourng greater tme coverage and data relablty over wder ndustry coverage. 4.1 Shft-share n detal In aggregate Between 1978 and 211, aggregate labour productvty n New Zealand grew by an average of 2.73% a year. 17 Ths can be decomposed nto a wthn-ndustry productvty ncrease of 3.2 percentage ponts (111% of the total growth of 2.73% a year), a small structural change level effect of.4 percentage ponts (1.6% of total growth) and a structural change growth effect of -.34 percentage ponts (-12% of total growth) (Fgure 13). The structural change level effect made a small postve contrbuton n most productvty growth cycles over the perod. However, t made a szable negatve contrbuton to productvty growth n the relatvely short 1997 to 2 cycle, and a neglgble negatve contrbuton from 199 to 1997 (Fgure 13). The structural change growth effect made a negatve contrbuton to productvty growth over most cycles, and was partcularly szable n the 199s (Fgure 13). Ths ndcates that ndustres wth faster labour productvty growth generally requred less labour nput per unt of output and that ther output growth was not ncreasng at a suffcent pace to offset the lower per unt labour nput needs. 17 The labour productvty growth rates presented here wll not necessarly match the offcal Statstcs New Zealand growth rates. See Appendx C for detals.